US8559722B2 - Image recognition apparatus and method - Google Patents

Image recognition apparatus and method Download PDF

Info

Publication number
US8559722B2
US8559722B2 US13/020,438 US201113020438A US8559722B2 US 8559722 B2 US8559722 B2 US 8559722B2 US 201113020438 A US201113020438 A US 201113020438A US 8559722 B2 US8559722 B2 US 8559722B2
Authority
US
United States
Prior art keywords
image
recognition
registered
image information
input
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active, expires
Application number
US13/020,438
Other languages
English (en)
Other versions
US20110194732A1 (en
Inventor
Ryosuke Tsuji
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Canon Inc
Original Assignee
Canon Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Canon Inc filed Critical Canon Inc
Assigned to CANON KABUSHIKI KAISHA reassignment CANON KABUSHIKI KAISHA ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: Tsuji, Ryosuke
Publication of US20110194732A1 publication Critical patent/US20110194732A1/en
Application granted granted Critical
Publication of US8559722B2 publication Critical patent/US8559722B2/en
Active legal-status Critical Current
Adjusted expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/172—Classification, e.g. identification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T1/00—General purpose image data processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/98—Detection or correction of errors, e.g. by rescanning the pattern or by human intervention; Evaluation of the quality of the acquired patterns
    • G06V10/993—Evaluation of the quality of the acquired pattern

Definitions

  • the present invention relates to an image recognition apparatus and method suited to an image capturing apparatus such as a video camera.
  • an image capturing apparatus having an object recognition function of recognizing a specific object contained in an image With the object recognition function, the image capturing apparatus can control the focus and exposure preferentially for a recognized object.
  • the image capturing apparatus having the object recognition function first detects an object image region containing a specific object from image data obtained by object recognition processing by the image capturing apparatus. Then, the image capturing apparatus extracts feature information from the image data of the detected object image region to identify the object region as a recognition target.
  • the image capturing apparatus compares the extracted feature information with that of an image registered in advance in a database. Based on the similarity obtained by the comparison, the image capturing apparatus recognizes a registered target which corresponds to the recognition target, or determines whether no corresponding registered target exists.
  • Japanese Patent Laid-Open No. 2003-187352 proposes a system which performs personal recognition to detect a person's face as a specific object region, extracts feature information of the face, and determines the person having this face.
  • processing of identifying the person (object) corresponding to a detected image such as a face will be called “recognition”.
  • the similarity is calculated by comparing a recognition target with a target registered in advance in a database.
  • a predetermined threshold is set for similarity to be calculated. If the calculated similarity is equal to or higher than the predetermined threshold, it is determined that the recognition target coincides with a target in the registration database.
  • the similarity is high.
  • the recognition target and the target registered in the database represent the same object, if the object status such as the object direction or brightness has changed, feature amounts (i.e. characteristics of the object and target that are compared) obtained from the target images differ from each other, and the similarity becomes low.
  • the similarity becomes lower than the predetermined threshold because (1) the recognition target does not coincide with a target registered in the database, or (2) the status of a target registered in the database does not coincide with that of an object serving as the recognition target. That is, it cannot be determined from only the similarity value that the recognition target is an object not registered in the database.
  • an embodiment of the present invention provides an image recognition apparatus and method intending more accurately to determine that an object detected from an image has not been registered in a database.
  • an image recognition apparatus configured to recognize an object in an input image
  • the apparatus comprising: a database in which an image or an image feature obtained by imaging an object, and an object recognizability state of the object are registered as registered image information in association with an identification of the object; a first calculation unit configured to calculate a coincidence degree between an object recognizability state of the object image and an object recognizability state in the registered image information; a second calculation unit configured to calculate a similarity between an image feature of the input object image and an image feature in the registered image information; and a recognition unit configured to execute recognition processing to recognize, based on the similarity and the coincidence degree, whether an object of the input object image is the same object as the object of the registered image information, wherein when the second calculation unit calculates that the similarity is lower than a first threshold and the first calculation unit calculates that the coincidence degree is equal to or higher than a second threshold, the recognition unit is configured to recognize that the object of the input object image is different from the object of the registered image information
  • a method of controlling an image recognition apparatus which recognizes an object in an input image comprising: a registration step of registering an image or an image feature obtained by imaging an object, and an object recognisability state regarding the object, as registered image information in a database in association with an identification of the object; a first calculation step of calculating a coincidence degree between an object recognisability state of the object image and an the object recognisability state in the registered image information; a second calculation step of calculating a similarity between an image or image feature of the object image and an image feature in the registered image information; and a recognition step of executing recognition processing to recognize, based on the similarity and the coincidence degree, whether an object of the object image is the same object as the object of the registered image information, wherein, when it is calculated that the similarity is lower than a first threshold and the coincidence degree is equal to or higher than a second threshold, the object of the object image is recognized in the recognition step to be different from the object of the registered
  • FIG. 1 is a block diagram exemplifying the arrangement of an image capturing apparatus according to the first embodiment
  • FIG. 2 is a flowchart showing recognition processing according to the first embodiment
  • FIG. 3 is a diagram exemplifying the recognition processing according to the first embodiment
  • FIG. 4 is a block diagram exemplifying the arrangement of an image capturing apparatus according to the second embodiment
  • FIG. 5 is a flowchart showing recognition processing according to the second embodiment.
  • FIG. 6 is a diagram exemplifying the recognition processing according to the second embodiment.
  • FIG. 1 is a block diagram showing the schematic arrangement of an image capturing apparatus 100 having an image recognition apparatus according to the first embodiment.
  • the image capturing apparatus 100 has a function of performing face-recognition to detect a person's face as a specific object region from the image data obtained by image sensing, extract feature information of the face, and identify the person having this face.
  • a light beam representing an object image converges through an imaging lens 101 , and enters an image sensing device 102 such as a CCD (charge-coupled device) image sensor or CMOS (complementary metal-oxide semiconductor) image sensor.
  • the image sensing device 102 outputs an electrical signal for each pixel in accordance with the intensity of the incident beam. This electrical signal is a video signal.
  • An analog signal processing unit 103 performs analog signal processing such as correlated double sampling (CDS) for the video signal output from the image sensing device 102 .
  • CDS correlated double sampling
  • An A/D conversion unit 104 converts the video signal output from the analog signal processing unit 103 into digital data, generating a digital image.
  • the digital image is input as an image to be processed to an imaging control unit 105 and image processing unit 106 .
  • the image processing unit 106 executes image processing including gamma correction and white balance processing for the input image to be processed.
  • the image processing unit 106 has a function of performing image processing using information about a specific object region in an image that is supplied from an object detection unit 109 and an object recognition unit 111 (to be described later).
  • the video signal output from the image processing unit 106 is sent to a display unit 107 .
  • the display unit 107 is, for example, an LCD (liquid crystal display) or organic EL (electroluminescent) display, and displays a video signal.
  • the display unit 107 can function as an electronic viewfinder (EVF) to sequentially display images detected in time series.
  • EVF electronic viewfinder
  • the video signal is recorded on a recording medium 108 (for example, a removable memory card).
  • the recording destination may be the built-in memory of a camera, a memory inserted into the camera, or a communicably connected external device.
  • the video signal (input image to be processed) output from the image processing unit 106 is also supplied to the object detection unit 109 .
  • the object detection unit 109 detects a specific object in the image, and specifies the number of objects and the object region.
  • the target object in the embodiment may be a person's face.
  • the detection method can thus be a known face detection method.
  • Known face detection techniques include, for example, a method using knowledge (skin color information and parts such as the eye, nose, and mouth) about the face, and a method which configures an identification unit for face detection based on a learning algorithm typified by the neural net. For high detection precision, a plurality of face detection methods are generally combined to perform face detection.
  • Information output from the object detection unit 109 includes object positions, sizes, tilts, detection result reliabilities, and the like by the number of detected objects.
  • a status determination unit 110 accesses a database 112 to determine whether the status (i.e. a recognisability) of the recognition target in the object recognition unit 111 (to be described later) and that of a target registered in advance in the database 112 coincide with each other, or determine the coincidence degree. Details of the status determination unit 110 will be described later.
  • the object recognition unit 111 extracts feature information from the image data (object image) of the object detected by the object detection unit 109 to identify the object region as a recognition target.
  • the object recognition unit 111 compares the extracted feature information with the feature information of an object image registered in advance in the database 112 . More specifically, the object recognition unit 111 compares the image feature(s) of the object image with that of an object registered in the database 112 . For example, by using information about face parts extracted upon face detection, the object recognition unit 111 obtains values for a plurality of types of items such as the distance between the centers (e.g.
  • the object recognition unit 111 normalizes the value of each item so that the value of any item (for example, the distance between the centers of the right and left eyes) coincides with a predetermined value.
  • the object recognition unit 111 converts the value of each item into points using a table prepared in advance.
  • the object recognition unit 111 calculates the ratio of the points of different items for various combinations of items, including, for example, the ratio of the points of the average of the sizes of the right and left eyes to those of the distance between the centers of the right and left eyes, and the ratio of the points of the distance from the bisecting point between the centers of the right and left eyes to the tip of the chin to those of the width of the mouth.
  • the object recognition unit 111 reads out ratio data of a person out of those of persons registered in advance in the database 112 , and obtains the percentage (%) of each calculated ratio with respect to the readout ratio data.
  • the object recognition unit 111 obtains the average of percentages (%) calculated for various combinations of items, and sets it as a recognition evaluation value for the person.
  • the object recognition unit 111 Based on a recognition evaluation value indicating the similarity obtained by the comparison, the object recognition unit 111 recognizes a registered target which corresponds to the recognition target, or determines whether no corresponding registered target exists. For example, a predetermined threshold is set for the recognition evaluation value. When the obtained recognition evaluation value is equal to or larger than the predetermined threshold, it can be determined that the recognition target is a target registered in the database 112 . The recognition evaluation value becomes smaller than the predetermined threshold in the following two cases:
  • the recognition target is different from a target registered in the database 112 , or
  • the recognition target coincides with a target in the registration database, but a recognition evaluation value that is equal to or larger than the predetermined threshold cannot be obtained.
  • the phenomenon in which the recognition target coincides with a target registered in the database 112 , but a recognition evaluation value equal to or larger than the predetermined threshold cannot be obtained occurs when, for example, the status (for example, face direction) of the object in the image is different. In other words, the input object is not recognizable as being similar to the target registered in the database. For this reason, when the recognition evaluation value is smaller than the predetermined threshold and the status determination unit 110 determines that the statuses of the two targets coincide with each other (or the status coincidence degree is equal to or higher than a predetermined threshold, meaning that the recognizability of the object is similar enough that if they were the same object, this would be detected), the object recognition unit 111 determines that the recognition target is an object different from that in the database. In other words, it is not simply because one of the faces being recognized is turned away that the object recognition unit 111 is not finding the faces of the input and of the database as being similar. Rather, the fact that the faces are actually different is more reliably concluded.
  • the status in the status determination unit 110 is defined by factors which decrease the precision of the object recognition unit 111 .
  • the factors which decrease the precision of the object recognition unit 111 depend on the algorithm of the object recognition unit 111 and the like, so the status in the status determination unit 110 is defined based on the characteristics of the object recognition unit 111 .
  • Examples of the status or recognisability state are the face direction, brightness, expression, concealment degree (for example, whether an object wears spectacles or a beard), and object region size.
  • a target object image, or the image feature amount and object recognisability state for object recognition are registered in association with the object.
  • object recognisability state what is meant is that a composition of the image is registered. This composition may include a direction in which a person is facing (i.e. directly at the camera capturing the image or to one side or in profile).
  • the object recognisability state may alternatively indicate if the object being imaged is partially obscured. In other words, the object recognisability state is the same as the object “status” described above.
  • a plurality of objects may be registered in the database 112 . Also, a plurality of data can be registered for the same object in the database.
  • the object can be registered with a plurality of different statuses.
  • an object recognition image may be registered in the database 112 .
  • the image feature amount and object recognisability state or the image registered in association with the object in the database 112 will be generically called “registered image information”.
  • the imaging control unit 105 controls the focus control mechanism and exposure control mechanism (neither is shown) of the imaging lens based on the video signal output from the A/D conversion unit 104 .
  • the imaging control unit 105 can use information of the target object region extraction result that is supplied from the object detection unit 109 and the object recognition unit 111 .
  • the image capturing apparatus 100 has a function of performing image sensing processing considering information of a specific object region in a captured image.
  • the imaging control unit 105 also controls the output timing, output pixel, and the like of the image sensing device 102 .
  • the image processing unit 106 reads, as an input image, an image obtained by digitizing, by the A/D conversion unit 104 , a captured image obtained via the image sensing device 102 (step S 201 ).
  • the object detection unit 109 detects an object region in the input image (step S 202 ).
  • the status determination unit 110 and the object recognition unit 111 obtain the image of the object region detected by the object detection unit 109 as an object image to be recognized.
  • the number of detection results, the position and size of each detection result in the image, the reliability of the detection result, and the like are obtained.
  • the status determination unit 110 and object recognition unit 111 execute processes in steps S 203 to S 209 for each object extracted as a recognition target by the object detection unit 109 . It is also possible to select a main object from a plurality of objects extracted by the object detection unit 109 , and to perform the processes in steps S 203 to S 209 for only the selected object as a recognition target.
  • the main object to be selected is, for example, an object having the largest evaluation value or an object having the largest area of the object region in detection of the object region (object image).
  • the status determination unit 110 calculates the status coincidence degree to determine the degree at which the status of the object extracted by the object detection unit 109 coincides with that of an object registered in the database 112 (step S 203 (first calculation processing)).
  • a higher status coincidence degree means higher status similarity.
  • An example of the status coincidence determination element is the face direction.
  • the feature points of face parts such as the eye and the mouth are extracted from the image of a recognition target and that of a target registered in the database, and face directions are estimated from the geometrical relationships between the extracted feature points.
  • the estimated face direction of the recognition target (i.e. the object image to be recognized) and that of the target (i.e. the object) registered in the database are compared, thereby determining the status coincidence degree.
  • the status in status coincidence determination suffices to be a factor which influences the precision of the object recognition unit 111 , and is not limited to face direction.
  • the status used here is defined in accordance with the system and algorithm of the object recognition unit 111 , and includes “brightness”, “expression”, “concealment degree”, and “size” in addition to face direction.
  • a method of determining the coincidence degree of each status will be exemplified.
  • the average of luminance values in the image of a recognition target and that of luminance values in the image of a target registered in the database are calculated, and in the case where the difference between the averages is 20% or less of an average calculated from a database image, it is determined that the luminance statuses coincide with each other, and a higher status coincidence degree is given for a lower percentage.
  • the reference value of 20% may be changed to a stricter (i.e. lower) value for a lower database image luminance.
  • expression the feature points of face parts such as the eye and the mouth are extracted from the image of a recognition target and that of a target registered in the database, and the facial expression in each image is determined from the degree of opening of the eye (eye height with respect to the eye width), the mouth shape (whether the two corners of the mouth are higher than its center), the position of the corner of the eye, and the like.
  • concealing features such as a mask and spectacles are detected from the image of a recognition target and that of a target registered in the database, and the difference between the ratios of the area of concealing features (or, indeed, the concealed portions of the object) detected in respective images to the area of the object (such as a face) is 5% or less of the ratio of the area of a concealing feature in a database image with respect to the object (i.e.
  • the object recognition unit 111 compares the image feature of the object image extracted by the object detection unit 109 with that of an object registered in the database 112 , calculating the recognition evaluation value in the second calculation processing (step S 204 ).
  • the recognition evaluation value indicates the similarity between the object image to be recognized and the object registered in the database 112 .
  • a larger recognition evaluation value means that the recognition target and the object registered in the database 112 are more highly likely to coincide with each other.
  • the object recognition unit 111 compares the recognition evaluation value with the first threshold, and if the recognition evaluation value is equal to or larger than the first threshold, determines that the recognition target is an object coincident with one registered in the database 112 (YES in step S 205 ⁇ step S 206 ). If the recognition evaluation value is smaller than the first threshold (NO in S 205 ), the object recognition unit 111 compares the status coincidence degree with the second threshold. If the status coincidence degree is equal to or higher than the second threshold, that is, the status of the object to be recognized coincides with that of the object registered in the database, the object recognition unit 111 determines that the recognition target is an object different from one registered in the database 112 (YES in step S 207 ⁇ step S 208 ).
  • step S 207 If the status coincidence degree is lower than the predetermined threshold (NO in step S 207 ), the object recognition unit 111 determines that the determination is impossible, without determining whether the recognition target differs from or coincides with the object in the database, and the determination becomes impossible (NO in step S 207 ⁇ step S 209 ).
  • the status determination unit 110 determines the coincidence degree of the face direction.
  • reference numeral 301 denotes an image registered in the database; and 302 to 304 , input images.
  • the recognition evaluation value of the input image 302 with the registered image 301 is equal to or larger than the first threshold, so it is determined that the recognition target is an object identical to one in the database.
  • the recognition evaluation value of the input image 303 with the registered image 301 is smaller than the first threshold, the status coincidence degree is equal to or higher than the second threshold (i.e.
  • both images have faces facing the same way, so a difference in recognition evaluation value is not because of a difference in status), and thus it is determined that the recognition target is an object different from a target registered in the database.
  • the recognition evaluation value of the input image 304 with the registered image 301 is smaller than the first threshold, the status coincidence degree is lower than the second threshold, and thus it is not determined whether the recognition target is an object identical to or different from a target registered in the database. In other words, determination of whether the object from the input image is the same as the stored object image is impossible.
  • a recognition evaluation value obtained by object recognition, and a status coincidence degree regarding the object status are obtained for an object image to be recognized when compared with an object registered in the database.
  • the recognition evaluation value is smaller than a predetermined threshold, it is determined by referring to the status coincidence degree whether the recognition target is different from the registered object. In other words, it can be more accurately determined that the recognition target is an object not registered in the object recognition database.
  • step S 207 determination is made using only the status coincidence degree, but the present invention is not limited to this.
  • the recognition evaluation value obtained in step S 204 is smaller than a third threshold (where the third threshold is smaller than the first threshold), and the status coincidence degree is higher than the second threshold (S 207 )
  • the process may advance from step S 205 to step S 206 only when the recognition evaluation value is equal to or larger than the first threshold and the status coincidence degree is equal to or higher than a fourth threshold, the fourth threshold being the same as or different from the second threshold, but included in an embodiment where both S 205 and S 207 must be satisfied for the result of S 206 to be positive.
  • a fourth threshold being the same as or different from the second threshold, but included in an embodiment where both S 205 and S 207 must be satisfied for the result of S 206 to be positive.
  • FIG. 4 is a block diagram showing an image capturing apparatus according to the second embodiment. A difference of the arrangement and operation of an image capturing apparatus 100 according to the second embodiment from those in the first embodiment will be mainly explained.
  • the same reference numerals as those in FIG. 1 denote units which have the same arrangements and execute the same operations and processes as those in the first embodiment.
  • An object tracking unit 401 tracks the same object between successive frames based on position and size similarities between detection results supplied from an object detection unit 109 for the successive frames. By specifying the same object in the moving image, the past result of an object recognition unit 111 for the object can be held and applied.
  • An execution determination unit 402 determines whether the object recognition unit 111 is to execute recognition processing for each object extracted by the object detection unit 109 . Also, the execution determination unit 402 selects an object to undergo recognition processing for the recognition target, from objects registered in a database 112 . Arranging the execution determination unit 402 as shown in FIG. 4 allows the use of the result of a status determination unit 110 and the past result of the object recognition unit 111 , and the selection of an object to undergo recognition processing (by the object recognition unit 111 ) from objects registered in the database 112 .
  • the object tracking unit 401 can specify the same object in a moving image which is continuous in time series, the past result of the object recognition unit 111 can be inherited, omitting further recognition processing by the object recognition unit 111 . That is, when the recognition evaluation value satisfies a predetermined threshold and it is determined that the object has been registered in the database 112 , this determination can be inherited by specifying the object by the object tracking unit 401 in subsequent frames.
  • recognition processing with the object registered in the database 112 can be omitted for the same recognition target in consecutive frames. Assuming that a plurality of types of image features and statuses are registered for the same object (i.e. an object with the same identification) in the database 112 , and it is determined from collation with one of the image features and/or statuses that the recognition target is different from the registered object.
  • recognition processing with the recognition target can be omitted for all the image features and statuses registered in the database 112 in regard to the object. Further, when an object to undergo recognition processing by the object recognition unit 111 does not remain, it can be determined that the recognition target is an object not registered in the database.
  • the object detection unit 109 obtains an object region (object image) from an input image obtained by digitizing a captured image.
  • Steps S 503 to S 513 are executed for each object extracted by the object detection unit 109 .
  • the object tracking unit 401 tracks a same object between frames which are successive in time series (step S 503 ). More specifically, if an object—whose position and size similarities between the successive frames satisfy predetermined criteria—exists among objects detected by the object detection unit 109 , the object tracking unit 401 determines that an identical object exists. If an object—whose position and size similarities between successive frames satisfy predetermined criteria—does not exist, the object tracking unit 401 regards the object as one which newly appears.
  • the execution determination unit 402 determines, for the object (recognition target) determined by the object tracking unit 401 not to be a new object, whether it has been determined in a past frame that the object coincides with a specific object registered in the database 112 , or whether it has been determined in a past frame that the object has not been registered in the database (whether it has been determined that the object is different from all objects registered in the database 112 ) (step S 504 ). If it has been determined that the recognition target coincides with a specific object, or that the recognition target has not been registered in the database, the execution determination unit 402 returns the process to step S 501 (YES in step S 504 ) to skip determination of the status coincidence degree and object recognition processing for the object.
  • the object recognition unit 111 determines whether registered image information to undergo object recognition for the recognition target remains in the database 112 (NO in step S 504 ⁇ step S 505 ). If registered image information to undergo object recognition does not remain in the database 112 , the recognition target is different from all objects registered in the database 112 . Thus, the object recognition unit 111 determines that the recognition target has not been registered in the database (NO in step S 505 ⁇ step S 506 ). After that, the process returns to step S 501 .
  • steps S 507 to S 513 are executed.
  • steps S 507 to S 513 are executed for all images registered in the database 112 that have been determined in step S 505 (YES in S 505 ) to undergo object recognition for the recognition target.
  • the status determination unit 110 determines the status coincidence degree of the recognition target to undergo object recognition (step S 507 ).
  • the processes in steps S 507 to S 513 are the same as those in steps S 203 to S 209 of FIG. 2 . Through the processes in steps S 507 to S 513 , one of the following results is obtained:
  • the status determination unit 110 determines the coincidence degree of the face direction.
  • reference numerals 601 to 603 denote registered images of objects registered in the database.
  • the registered images 601 and 602 represent identical objects (target A) in different face directions and are given the identifications (ID) A- 1 and A- 2 respectively.
  • the registered image 603 represents an object (target B with ID B- 1 ) different from the object of the registered images 601 and 602 .
  • Reference numerals 604 to 607 denote input images in scene 1 ; and 608 to 611 , input images in scene 2 .
  • Reference numeral 612 denotes an object recognition execution result table for scene 1 ; and 613 , an object recognition execution result table for scene 2 .
  • the execution result tables 612 and 613 are updated in steps S 510 , S 512 and S 513 .
  • the recognition evaluation value of the recognition target for an image with the database ID: A- 1 is equal to or larger than a threshold, so it is determined that the recognition target coincides with a target in the database.
  • recognition processing for image with database ID: A- 2 the recognition evaluation value is smaller than the threshold, the status coincidence degree is lower than a threshold, and it is determined that the determination is impossible.
  • recognition processing for the image with database ID: B- 1 the recognition evaluation value is smaller than the threshold, the status coincidence degree is equal to or higher than the threshold, and it is determined that the recognition target is an object different from target B.
  • the first input image 604 has been determined as being the same as the registered image with ID: A- 1 and so this reference can be used for the rest of the frames 2 to 4 where the same face is imaged.
  • the recognition target is an object different from target A indicated by the database ID: A- 2 because the recognition evaluation value is smaller than the threshold and the status coincidence degree is equal to or higher than the threshold.
  • This result that the recognition target is not target A
  • the database ID: A- 1 and ID: A- 2 in frame 3 neither object recognition nor determination of the status coincidence degree is executed for the database ID: A- 1 and ID: A- 2 in frame 3 .
  • object recognition and determination of the status coincidence degree are executed only for the database ID: B- 1 in frame 3 of the input image.
  • the recognition target is an object different from target B indicated by the database ID: B- 1 because the recognition evaluation value for the database ID: B- 1 is smaller than the threshold and the status coincidence degree is equal to or higher than the threshold. Since it is determined that the recognition target in scene 2 is different from all targets (targets A and B in this example) in the database, it is determined that the recognition target has not been registered in the database (step S 506 ). In frame 4 and subsequent frames, neither object recognition nor determination of the status coincidence degree is executed for the object, as the object in the frames is known not to be one of those in the database.
  • a determination result obtained in a past frame by object recognition and determination of the status coincidence degree for the identical recognition target is inherited, increasing the processing efficiency.
  • the processing can be made more efficient based on the result of the status determination unit 110 .
  • the object recognition unit 111 is less likely to obtain a recognition evaluation value which satisfies a predetermined threshold.
  • processing by the object recognition unit 111 based on the database may be omitted for the frame.
  • a person's face is detected as an object.
  • the present invention is not limited to a person's face, and is applicable to even the face of an animal such as a pet.
  • the present invention is not limited to the face, and is applicable to a person, animal, automobile, and the like as long as object recognition is executed to extract and specify a target object.
  • the above embodiments have described an image capturing apparatus, but the present invention is not limited to the image capturing apparatus and is applicable to even a reproduction apparatus such as a printer or display screen.
  • the present invention can be applied to the reproduction apparatus by using reproduction data as data of object recognition processing.
  • the above-described embodiments can more accurately determine that an object detected from an image has not been registered in a database compared with simply comparing an input image with a database image.
  • aspects of the present invention can also be realized by a computer of a system or apparatus (or devices such as a CPU (central processing unit) or MPU (microprocessing unit) that reads out and executes a program recorded on a memory device to perform the functions of the above-described embodiment(s), and by a method, the steps of which are performed by a computer of a system or apparatus by, for example, reading out and executing a program recorded on a memory device to perform the functions of the above-described embodiment(s).
  • the program is provided to the computer for example via a network or from a recording medium of various types serving as the memory device (for example, computer-readable storage medium).
US13/020,438 2010-02-10 2011-02-03 Image recognition apparatus and method Active 2032-05-04 US8559722B2 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2010-028207 2010-02-10
JP2010028207A JP5567853B2 (ja) 2010-02-10 2010-02-10 画像認識装置および方法

Publications (2)

Publication Number Publication Date
US20110194732A1 US20110194732A1 (en) 2011-08-11
US8559722B2 true US8559722B2 (en) 2013-10-15

Family

ID=44226083

Family Applications (1)

Application Number Title Priority Date Filing Date
US13/020,438 Active 2032-05-04 US8559722B2 (en) 2010-02-10 2011-02-03 Image recognition apparatus and method

Country Status (5)

Country Link
US (1) US8559722B2 (zh)
EP (1) EP2357589B1 (zh)
JP (1) JP5567853B2 (zh)
KR (1) KR101280920B1 (zh)
CN (2) CN102147856B (zh)

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150310261A1 (en) * 2014-04-28 2015-10-29 Microsoft Corporation Creation of representative content based on facial analysis
US9477625B2 (en) 2014-06-13 2016-10-25 Microsoft Technology Licensing, Llc Reversible connector for accessory devices
US9614724B2 (en) 2014-04-21 2017-04-04 Microsoft Technology Licensing, Llc Session-based device configuration
US20170124428A1 (en) * 2015-11-04 2017-05-04 Samsung Electronics Co., Ltd. Authentication method and apparatus, and method and apparatus for training a recognizer
US9717006B2 (en) 2014-06-23 2017-07-25 Microsoft Technology Licensing, Llc Device quarantine in a wireless network
US9773156B2 (en) 2014-04-29 2017-09-26 Microsoft Technology Licensing, Llc Grouping and ranking images based on facial recognition data
US9874914B2 (en) 2014-05-19 2018-01-23 Microsoft Technology Licensing, Llc Power management contracts for accessory devices
US10111099B2 (en) 2014-05-12 2018-10-23 Microsoft Technology Licensing, Llc Distributing content in managed wireless distribution networks
US10691445B2 (en) 2014-06-03 2020-06-23 Microsoft Technology Licensing, Llc Isolating a portion of an online computing service for testing
US11361547B2 (en) 2017-12-08 2022-06-14 Nec Communication Systems, Ltd. Object detection apparatus, prediction model generation apparatus, object detection method, and program

Families Citing this family (39)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2405393B1 (en) * 2009-03-04 2021-07-28 Nec Corporation Device, method and program for creating information for object position estimation
JP5918996B2 (ja) * 2011-12-27 2016-05-18 キヤノン株式会社 被写体認識装置および辞書データ登録方法
US20150088625A1 (en) * 2012-01-30 2015-03-26 Nokia Corporation Method, an apparatus and a computer program for promoting the apparatus
JP5939705B2 (ja) * 2012-02-08 2016-06-22 カシオ計算機株式会社 被写体判定装置、被写体判定方法及びプログラム
JP6003124B2 (ja) * 2012-03-15 2016-10-05 オムロン株式会社 認証装置、認証装置の制御方法、制御プログラム、および記録媒体
JP5959923B2 (ja) * 2012-04-26 2016-08-02 キヤノン株式会社 検出装置、その制御方法、および制御プログラム、並びに撮像装置および表示装置
US20140036152A1 (en) * 2012-07-31 2014-02-06 Google Inc. Video alerts
US8948568B2 (en) 2012-07-31 2015-02-03 Google Inc. Customized video
CN103902967A (zh) * 2012-12-27 2014-07-02 杭州美盛红外光电技术有限公司 图像识别控制装置和图像识别控制方法
JP6025557B2 (ja) * 2012-12-27 2016-11-16 キヤノン株式会社 画像認識装置、その制御方法及びプログラム
WO2014101803A1 (zh) * 2012-12-27 2014-07-03 Wang Hao 红外选择装置和红外选择方法
CN103900711A (zh) * 2012-12-27 2014-07-02 杭州美盛红外光电技术有限公司 红外挑选装置和红外挑选方法
CN103900712A (zh) * 2012-12-27 2014-07-02 杭州美盛红外光电技术有限公司 热像选择装置和热像选择方法
CN103900710A (zh) * 2012-12-27 2014-07-02 杭州美盛红外光电技术有限公司 热像挑选装置和热像挑选方法
WO2014101809A1 (zh) * 2012-12-27 2014-07-03 Wang Hao 热像检测装置和热像检测方法
US9201499B1 (en) * 2013-02-11 2015-12-01 Amazon Technologies, Inc. Object tracking in a 3-dimensional environment
US9501684B1 (en) * 2013-12-13 2016-11-22 Google Inc. Providing non-destructive editing reconciliation
JP6362085B2 (ja) * 2014-05-21 2018-07-25 キヤノン株式会社 画像認識システム、画像認識方法およびプログラム
US9934563B2 (en) 2015-10-29 2018-04-03 Empire Technology Development Llc 3D object rotation-based mechanical parts selection through 2D image processing
US10546185B2 (en) * 2015-12-01 2020-01-28 Casio Computer Co., Ltd. Image processing apparatus for performing image processing according to privacy level
JP6378664B2 (ja) * 2015-12-02 2018-08-22 株式会社オプティム 情報特定システム、情報特定方法及びそのプログラム
US10482681B2 (en) 2016-02-09 2019-11-19 Intel Corporation Recognition-based object segmentation of a 3-dimensional image
US10373380B2 (en) 2016-02-18 2019-08-06 Intel Corporation 3-dimensional scene analysis for augmented reality operations
US10359806B2 (en) * 2016-03-28 2019-07-23 Sony Interactive Entertainment Inc. Pressure sensing to identify fitness and comfort of virtual reality headset
CN105956658A (zh) * 2016-04-29 2016-09-21 北京比特大陆科技有限公司 数据处理方法、数据处理装置及芯片
CN105976027A (zh) * 2016-04-29 2016-09-28 北京比特大陆科技有限公司 数据处理方法和装置、芯片
JP6840478B2 (ja) * 2016-07-07 2021-03-10 キヤノン株式会社 電子機器
US10573018B2 (en) * 2016-07-13 2020-02-25 Intel Corporation Three dimensional scene reconstruction based on contextual analysis
KR101855990B1 (ko) * 2016-10-31 2018-05-10 (주)노바코스 차량 통과 감지 장치 및 방법
CN109272006A (zh) * 2017-07-18 2019-01-25 北京柯斯元科技有限公司 随机纹理防伪标识图案判定系统及判定方法
CN109635703A (zh) * 2018-12-05 2019-04-16 成都牧云慧视科技有限公司 一种可间断登记并持续进行基准数据更新的识别方法
CN109886145B (zh) * 2019-01-29 2021-09-03 浙江泽曦科技有限公司 宠物脸部识别方法及系统
TWI704505B (zh) * 2019-05-13 2020-09-11 和碩聯合科技股份有限公司 人臉辨識系統、建立人臉辨識之資料之方法及其人臉辨識之方法
JP7235119B2 (ja) * 2019-08-08 2023-03-08 日本電気株式会社 処理装置、会計装置、処理方法及びプログラム
JP7293039B2 (ja) * 2019-08-16 2023-06-19 キヤノン株式会社 撮像装置およびその制御方法
CN110928216B (zh) * 2019-11-14 2020-12-15 深圳云天励飞技术有限公司 人工智能装置
CN111696184B (zh) * 2020-06-10 2023-08-29 上海米哈游天命科技有限公司 骨骼蒙皮融合确定方法、装置、设备和存储介质
KR102464197B1 (ko) * 2022-05-25 2022-11-09 주식회사 유투에스알 빅데이터 기반 영상 감시 방법
KR102464196B1 (ko) * 2022-05-25 2022-11-09 주식회사 유투에스알 빅데이터 기반 영상 감시시스템

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5450504A (en) * 1992-05-19 1995-09-12 Calia; James Method for finding a most likely matching of a target facial image in a data base of facial images
US20020103574A1 (en) 2001-01-30 2002-08-01 Junichi Funada Robot, identifying environment determining method, and program thereof
JP2002251380A (ja) 2001-02-22 2002-09-06 Omron Corp 利用者照合システム
JP2003187352A (ja) 2001-12-14 2003-07-04 Nippon Signal Co Ltd:The 特定人物検出システム
US20090067726A1 (en) * 2006-07-31 2009-03-12 Berna Erol Computation of a recognizability score (quality predictor) for image retrieval
US20090309702A1 (en) 2008-06-16 2009-12-17 Canon Kabushiki Kaisha Personal authentication apparatus and personal authentication method

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005056004A (ja) * 2003-08-07 2005-03-03 Omron Corp 顔照合装置、顔照合方法、および顔照合プログラム
JP2005084815A (ja) * 2003-09-05 2005-03-31 Toshiba Corp 顔認識装置、顔認識方法および通行制御装置
US20060120571A1 (en) * 2004-12-03 2006-06-08 Tu Peter H System and method for passive face recognition
JP2006236255A (ja) * 2005-02-28 2006-09-07 Mitsubishi Electric Corp 人物追跡装置および人物追跡システム
JP4595750B2 (ja) * 2005-08-29 2010-12-08 ソニー株式会社 画像処理装置および方法、並びにプログラム
KR100745981B1 (ko) * 2006-01-13 2007-08-06 삼성전자주식회사 보상적 특징에 기반한 확장형 얼굴 인식 방법 및 장치
JP4845715B2 (ja) * 2006-12-22 2011-12-28 キヤノン株式会社 画像処理方法、画像処理装置、プログラム、及び記憶媒体
JP4952920B2 (ja) * 2007-06-04 2012-06-13 カシオ計算機株式会社 被写体判定装置、被写体判定方法及びそのプログラム
CN100481112C (zh) * 2007-06-11 2009-04-22 湖北东润科技有限公司 一种人脸识别系统与识别方法
CN100463000C (zh) * 2007-09-27 2009-02-18 上海交通大学 基于级联分类和哈夫圆变换的人眼状态检测方法
JP2009163555A (ja) * 2008-01-08 2009-07-23 Omron Corp 顔照合装置
US8180112B2 (en) * 2008-01-21 2012-05-15 Eastman Kodak Company Enabling persistent recognition of individuals in images
CN100557625C (zh) * 2008-04-18 2009-11-04 清华大学 人脸部件特征和Gabor人脸特征融合的人脸识别方法及其装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5450504A (en) * 1992-05-19 1995-09-12 Calia; James Method for finding a most likely matching of a target facial image in a data base of facial images
US20020103574A1 (en) 2001-01-30 2002-08-01 Junichi Funada Robot, identifying environment determining method, and program thereof
JP2002251380A (ja) 2001-02-22 2002-09-06 Omron Corp 利用者照合システム
JP2003187352A (ja) 2001-12-14 2003-07-04 Nippon Signal Co Ltd:The 特定人物検出システム
US20090067726A1 (en) * 2006-07-31 2009-03-12 Berna Erol Computation of a recognizability score (quality predictor) for image retrieval
US20090309702A1 (en) 2008-06-16 2009-12-17 Canon Kabushiki Kaisha Personal authentication apparatus and personal authentication method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Lee, et al. "Measurement of Face Recognizability for Visual Surveillance." ICAPR 2005, LNCS. 3687. (2005): 287-294. Print. *
The above references were cited in a Dec. 29, 2011 European Search Report which is enclosed of the counterpart European Patent Application No. 11152123.3.

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9614724B2 (en) 2014-04-21 2017-04-04 Microsoft Technology Licensing, Llc Session-based device configuration
US20150310261A1 (en) * 2014-04-28 2015-10-29 Microsoft Corporation Creation of representative content based on facial analysis
US10311284B2 (en) 2014-04-28 2019-06-04 Microsoft Technology Licensing, Llc Creation of representative content based on facial analysis
US9639742B2 (en) * 2014-04-28 2017-05-02 Microsoft Technology Licensing, Llc Creation of representative content based on facial analysis
US9773156B2 (en) 2014-04-29 2017-09-26 Microsoft Technology Licensing, Llc Grouping and ranking images based on facial recognition data
US10607062B2 (en) 2014-04-29 2020-03-31 Microsoft Technology Licensing, Llc Grouping and ranking images based on facial recognition data
US10111099B2 (en) 2014-05-12 2018-10-23 Microsoft Technology Licensing, Llc Distributing content in managed wireless distribution networks
US9874914B2 (en) 2014-05-19 2018-01-23 Microsoft Technology Licensing, Llc Power management contracts for accessory devices
US10691445B2 (en) 2014-06-03 2020-06-23 Microsoft Technology Licensing, Llc Isolating a portion of an online computing service for testing
US9477625B2 (en) 2014-06-13 2016-10-25 Microsoft Technology Licensing, Llc Reversible connector for accessory devices
US9717006B2 (en) 2014-06-23 2017-07-25 Microsoft Technology Licensing, Llc Device quarantine in a wireless network
US20170124428A1 (en) * 2015-11-04 2017-05-04 Samsung Electronics Co., Ltd. Authentication method and apparatus, and method and apparatus for training a recognizer
US10275684B2 (en) * 2015-11-04 2019-04-30 Samsung Electronics Co., Ltd. Authentication method and apparatus, and method and apparatus for training a recognizer
US11361547B2 (en) 2017-12-08 2022-06-14 Nec Communication Systems, Ltd. Object detection apparatus, prediction model generation apparatus, object detection method, and program

Also Published As

Publication number Publication date
KR101280920B1 (ko) 2013-07-02
CN103353933A (zh) 2013-10-16
KR20110093659A (ko) 2011-08-18
EP2357589B1 (en) 2019-06-12
CN102147856B (zh) 2013-06-05
EP2357589A3 (en) 2012-01-25
CN103353933B (zh) 2016-12-28
CN102147856A (zh) 2011-08-10
JP2011165008A (ja) 2011-08-25
US20110194732A1 (en) 2011-08-11
JP5567853B2 (ja) 2014-08-06
EP2357589A2 (en) 2011-08-17

Similar Documents

Publication Publication Date Title
US8559722B2 (en) Image recognition apparatus and method
US9098760B2 (en) Face recognizing apparatus and face recognizing method
US8416987B2 (en) Subject tracking apparatus and control method therefor, image capturing apparatus, and display apparatus
US9323979B2 (en) Face recognition performance using additional image features
US8314854B2 (en) Apparatus and method for image recognition of facial areas in photographic images from a digital camera
US8836777B2 (en) Automatic detection of vertical gaze using an embedded imaging device
US8526742B2 (en) Image processing apparatus, method, and program that classifies data of images
US9571740B2 (en) Image recognition device that recognizes specific object area, method of controlling the device, and storage medium, as well as image pickup apparatus, and display device
US9633284B2 (en) Image processing apparatus and image processing method of identifying object in image
JP6025557B2 (ja) 画像認識装置、その制御方法及びプログラム
JP6157165B2 (ja) 視線検出装置及び撮像装置
CN103379279A (zh) 被摄体区域检测装置及其方法、图像摄取装置及显示装置
JP5971712B2 (ja) 監視装置及び方法
CN107148237B (zh) 信息处理装置、信息处理方法和程序
JP4840978B2 (ja) 撮像装置及びその制御方法及びプログラム及び記憶媒体
US10140503B2 (en) Subject tracking apparatus, control method, image processing apparatus, and image pickup apparatus
JP5539565B2 (ja) 撮像装置及び被写体追跡方法
JP5995610B2 (ja) 被写体認識装置及びその制御方法、撮像装置、表示装置、並びにプログラム
Marciniak et al. Influence of pose angle on face recognition from very low resolution images
JP2013157675A (ja) 撮像装置及びその制御方法、プログラム、並びに記憶媒体
JP2010154374A (ja) 撮像装置および被写体追跡方法

Legal Events

Date Code Title Description
AS Assignment

Owner name: CANON KABUSHIKI KAISHA, JAPAN

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:TSUJI, RYOSUKE;REEL/FRAME:026249/0482

Effective date: 20110201

STCF Information on status: patent grant

Free format text: PATENTED CASE

FEPP Fee payment procedure

Free format text: PAYOR NUMBER ASSIGNED (ORIGINAL EVENT CODE: ASPN); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

FPAY Fee payment

Year of fee payment: 4

MAFP Maintenance fee payment

Free format text: PAYMENT OF MAINTENANCE FEE, 8TH YEAR, LARGE ENTITY (ORIGINAL EVENT CODE: M1552); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

Year of fee payment: 8